Speaker: Ming-Jun Lai, Professor, Dept. of Math, University of Georgia (For more information about the speaker, please visit Speaker’s Website – http://alpha.math.uga.edu/~mjlai/ )
Date: March 8, Friday Time: 3:30 – 4:30 pm
Venue: MS 427
Summary: Prof. Lai first provided an introduction to multivariate splines over triangulation/tetrahedralization which are piece-wise polynomial functions. He then explained how to evaluate, how to take a derivative, and how to integrate these functions as well as the smoothness conditions across interior edges of triangulation and interior faces of tetrahedralization. Prof. Lai further discussed how to use them for scattered data interpolation/fitting as well as numerical solution of linear and nonlinear partial differential equations, e.g. Navier-Stokes equations, diffusion and reaction equations, and Helmholtz equations. Finally, he extended how to define spline functions over polygonal partitions and their numerical solution of elliptic PDE. This session was part of the bi-weekly seminar series in March held by SIAM University of Calgary student chapter!
Data Science Day 2019 was organized by SIAM University of Calgary Student Chapter on February 16th, 2019 at Science Theaters 127 and 130 at the University of Calgary. The event comprised of seminars by a reputed panel of scholars from departments ranging from Computer Science to Healthcare Data, Biostatistics, Finance and Business. Below is an overview of each talk,
- Dr. Mahshid Marbouti – Department of Computer Science Overview: In this exciting time for the field of data science, data science is used to extract insights from large amounts of data. Some examples are Text Analytics, Image Processing and Recommender Systems. In this talk Dr. Marbouti reviewed the foundations of data science field. Then, he spoke about the difference between Deep Learning and Machine Learning and how we can use cloud computing services like Microsoft Azure and AWS to develop practical and smart data science solutions.
- Dr. Usman Alim – Department of Computer Science Overview: The visualization of static and time-varying tridimensional scientific phenomena is an active area of research. In this talk Dr. Alim provided an overview of data structures and techniques used to represent and visualize scalar, vector and tensor fields in 3D. He then focus on the interactive visualization and manipulation of multivariate datasets. Multivariate visualizations attempt to co-visualize several variables in the same three-dimensional space in an expressive and illustrative manner so as to reduce occlusion and visual clutter. Interaction techniques play a crucial role as they allow the user to interactively explore several variables without losing spatial reference. Several projects were showcased – ranging from interactive lens-based techniques on surfaces and immersive environments to entropy based methods used to quantify salience in vector fields.
- Dr. Quan Long – Department of Bioinformatics Overview: One of the ultimate goals in the field of genomics is to develop a genome-based phenotype predictor, in other words, a way to confidently predict biological phenomena from specific genetic variations. Empowered by the recent advancement of high-throughput technologies, in addition to genomes, researchers can assess transcriptome, methylome, proteome and etc., referred to as “in-between-omes” hereafter. When multi-scale in-between-omes are present, how to statistically integrate them in a polygenic phenotype predictor is not a trivial task. This is because, due to the problem of overfitting, more in-between-ones data may cause an increased complexity (of the predictive models) that offsets the benefit of the extra information, which ends up lower power. In this project, Dr. Long proposed a novel approach that leverages in-between-omes to reduce, instead of increase, the complexity of the model. In his method, the role of in-between-ones is to provide prior knowledge that helps group the genomic markers more effectively so that the genome-based predictor will have fewer coefficients to estimate, therefore be more powerful, thanks to the reduced model complexity. He then tested the approach using many benchmark datasets in plants and humans that have been extensively reanalyzed in the field. It is demonstrated that that this approach increases the power of polygenic prediction substantially for most complex traits. This research leads to a new path towards effective integration of multi-scale data with polygenic predictors along the line of the “less-is-more” principle in the circumstance of “more is less”.
- Dr. Alberto Nettle-Aguirre – Biostatistics Overview: In this talk, Dr. Aguirre shed some light into some ideas of what we have to be careful as statisticians and data scientists, and also an example of how we can play a role in defining the research question in such a way that we can gain more form the data at hand and collaborate better with health researchers.
- Dr. Thierry Chekouo – Department of Mathematics and Statistics Overview: In biomedical research, investigators are generating large, complex, multidimensional, and diverse datasets through new technologies developed in an age when sequencing a human genome is inexpensive. Various types of genomic, epigenomic, transcriptomic, proteomic and radiomic datasets with different sizes, formats, and structures have become available. In this talk, Dr. Thierry introduced different data types and some statistical techniques used in biomedical data science. He also covered integrative statistical and computational frameworks that acknowledge and exploit inherent complex structural relationships (between data types) for both biomarker discovery and clinical prediction to aid translational medicine. These approaches were illustrated using several biomedical case examples.
- Nima Safaian – Head of Trading Analytics at Cenovus,Calgary. Overview: Data is changing how commodities trading and trading analytics is performed by energy companies and trading organizations. Cloud computing, streaming real-time data and machine learning, allow traders a near real time analysis and prediction of the commodities movements across the world. The need for Data Science skills is more than ever in trading and risk management organizations. Dr. Safaian discussed how data science is helping commodities trading organization to develop and sustain their competitive edge.
- Thomas Holloway -Business Overview: Responsive AI is a venture-backed startup in direct-to-client wealth management, hybrid wealth technology for existing wealth businesses, and data-driven client intelligence for larger institutions (that have more data). Responsive’s asset management business is differentiated by streamlined user experience, industry-leading cost, and the implementation of data science research for optimized portfolio performance. Eschewing security selection as a source of value, Responsive clients own portfolios of Exchange Traded Funds (ETFs). Each ETF provides exposure to an asset class such as US stocks, foreign stocks, Government bonds, and High Yield Bonds. The Responsive investment process is to reconsider the weights monthly using signals from markets and the economy. Following Paul Samuelson, “The market is micro efficient, but macro inefficient”. For the asset classes that clients own, we are interested in relative returns for asset pairs, for example equity-versus-cash, and within equity, say Canada-versus-US. An infrastructure was built to test, select, and combine the most reasonable, profitable, consistent, and additive factors. Price-based factors tend to take the form of valuation, momentum, and mean reversion. In addition though, since recessions are such significant divers of relative returns, macroeconomic forecasting is valuable, particularly in combination with market factors. Markets are adaptive ecosystems of people (and now machines). The modelling challenge is therefore behavioral and biological because the allocation decisions of individuals and investment committees reflexively help determine subsequent returns. Everyone is looking at subsets of the same data, headlines, and prices. For the first time ever, wealth managers have the opportunity to generate insights from real-time client data. In the past, managers focused on investment products and portfolios because client data was out of reach. Now, with account aggregation and digital banking, it is possible to develop highly personalized and adaptive wealth strategies. Human behavior might not be logical, but it is predictable. The ability to build wealth hinges on one’s personality and daily habits. Understanding how much risk to take on – financially and psychologically – and what habits to keep or kill are at the core of a successful long-term wealth plan. At Responsive, their mission is to understand wealth behavior through science.
“By applying insights from the brain and behavioral sciences, as well as contemporary machine learning, we can uncover complex financial portraits and life events that enable advisers to help people when it matters most.” – Dr. Logan Grosenick, Responsive VP of Research.
- Janice B. Elliason – Business Overview: In this talk, Ms. Elliason focused on applications of business analytics in various businesses. She presented the results of research and applications in diverse areas such as Westjet, health care, government, and retail.
Date: February 6, 2019
Location: MS 431, Time: 12:00 pm – 1:00 pm
Abstract: Hawkes processes are class of stochastic counting processes, that have been applied in diverse areas, from earthquake modelling to financial analysis. They are point processes whose defining characteristic is that they ‘self-excite’, meaning that each arrival increases the rate of future arrivals for some period of time. Thus, the Hawkes process is a mathematical model for these ‘self-exciting’ processes, named after its creator Alan G. Hawkes (1971), and it is a non-Markovian extension of the Poisson process. Hawkes models are becoming more and more popular in the domains of high frequency finance and insurance. The talk was devoted to the Hawkes processes, their applications in finance and insurance, and consisted of three parts. The first part introduced the Hawkes processes and described their properties. The second part was devoted to the modelling of trading activities in high frequency finance, namely, modelling of limit order books, and to the justification of their modelling by presenting some numerical examples using real data. The third part focused on application of Hawkes processes in insurance.
Bio: Dr. Anatoliy Swishchuk is a Professor in financial mathematics at the Department of Mathematics and Statistics, University of Calgary, Calgary, Canada. He got his B.Sc. and M.Sc. degrees from Kiev State University, Kiev, Ukraine. He is a holder of two doctorate degrees in Mathematics and Physics (Ph. D. and D. Sc.) from the prestigious National Academy of Sciences of Ukraine (NASU), Kiev, Ukraine, and is a recipient of NASU award for young scientist with a gold medal for series of research publications in random evolutions and their applications. Dr. Swishchuk is a chair of finance and energy finance seminar ’Lunch at the Lab’ at the Department of Mathematics and Statistics, and Calgary Site Director of Postdoctoral Training Center in Stochastics (PTCS). He was a steering committee member of the Professional Risk Managers International Association (PRMIA), Canada (2006-2015), and is a steering committee member of Global Association of Risk Professionals (GARP), Canada (since 2015). His research areas include financial mathematics, random evolutions and their applications, biomathematics, stochastic calculus, and he serves on editorial boards for four research journals. He is the author of more than 150 publications, including 13 books and more than 100 articles in peer-reviewed journals. Recently he received a Peak Scholar award.
University of Calgary, SIAM student chapter conducted a Fall 2018 Biweekly series in Quantum Optics which was presented by Dr. Christoph Simon, Department of Physics and Astronomy on November 27, 2018. Dr. Simon described the many facets of quantum optics research through examples from his own career. He spoke about the theoretical and practical work demonstrating quantum-non locality, superposition and entanglement to a macroscopic level. He then discussed implementing a global quantum network, on the possible existence of photonic communication channels in the brain and super-resolution imaging.
Dr. Christoph Simon studied physics at the University of Vienna, obtained his master’s degree at Ecole Normale Superieure in Paris and did a Ph.D with Anton Zeilinger from the University of Vienna. He was a postdoc with Dirk Bouwmeester in Oxford and UC Santa Barbara. He became an Associate Professor at the University of Calgary in 2009 and a professor in 2016.
Two MITACS Information Sessions were organised by the SIAM University of Calgary Student Chapter on November 1st and November 8th, 2018 to provide internship opportunities for graduate students. Internships from a broad spectrum of engineering and science disciplines are available for interested students. The session speakers Mr. Angelo Nwigwe and Mr. Oba Harding, who are the current Directors of Business Development, Mitacs Alberta provided information on internships, exchange opportunities and other mitacs programs.
Title: From Geometric Modelling to Digital Earth
Date: Thursday, March 29, 2018
Time: 3:00 PM – 4:00 PM
Title: Analysis of Clustered Diagnostic Tests Subject to Misclassification Using Latent Variable Modelling
Speaker: Dr. Hua Shen, Department of Mathematics & Statistics, University of Calgary
Location: Mathematical Sciences MS 431
Date: Thursday, March 15, 2018
Time: 2:00 PM – 3:00 PM
Dr. Innanen introduced the problem of interest and a number of related challenges, and then he talked about some methods by which those challenges can be addressed. He finished his presentation by an example which was one of the recent projects that he had worked on. This seminar was appreciated by a number of graduate students and faculty members, mainly from the faculty of science.
Speaker: Dr. Kristopher Innanen, University of Calgary
Location: Mathematical Sciences 431
Date & Time: February 08, 2018. 2:00 pm – 3:00 pm
Title: Elastic Seismic FWI for Reservoir Characterization
Seismic full waveform inversion (FWI) as a technology is now regularly applied to problems in marine, i.e., offshore resource exploration and monitoring. FWI also seems to have the potential to become a major technology for monitoring elastic and rock physics properties of producing heavy oil and low-permeability reservoirs. However, some of the technical challenges faced by geophysicists in applying FWI, such as managing seismic attenuation, leakage between parameters, computational burden, etc., grow in land/elastic environments. I will give an overview of the look and feel of seismic FWI, which is at its heart a very large Newton optimization, point out where the technical bottlenecks are, discuss workarounds, and finish with an example of FWI applied in a producing Western Canadian unconventional reservoir that suggests we’re moving in the right direction.
About the speaker:
Kris Innanen received the BSc and MSc degrees from York University in 1996 and 1998, in Earth Science/Physics and Physics respectively, and the PhD degree in geophysics from the University of British Columbia in 2003. He joined the Department of Physics at the University of Houston as an Assistant Professor in 2005 and the Department of Geoscience at the University of Calgary as an Associate Professor in 2009. In 2016 he took on the directorship of the CREWES consortium. In 2006 he received the J. Clarence Karcher award from the Society of Exploration Geophysicists.